Yihao LIU

photo_lyh.jpg
Google scholar Github XiaoHongShu

I am a Research Scientist at the Shanghai Artificial Intelligence Laboratory, where I lead a team focusing on multimodal generation and understanding. I earned my Bachelor’s degree in 2018 and my Ph.D. in 2023, both from the University of Chinese Academy of Sciences (UCAS). During my doctoral studies, I was affiliated with the Shenzhen Institutes of Advanced Technology (SIAT), Chinese Academy of Sciences, under the supervision of Prof. Yu Qiao and Prof. Chao Dong. My research lies at the intersection of computer vision, generative modeling, and scientific intelligence, with particular emphasis on multimodal foundation models and image/video enhancement.

Throughout my student journey, I have been honored with prestigious awards, including the President’s Award of the Chinese Academy of Sciences, the Zhu Li Yue Hua Outstanding Doctoral Student Award, the CAS Excellent Youth League Member Award, the Beijing Outstanding Graduate Award, the SIAT President’s Innovation Award, as well as the CVMJ 2025 Best Paper Honorable Mention Award.

I have also excelled in multiple international and national competitions, such as 1st place in the PIRM 2018 Perceptual Image Super-Resolution Challenge, 1st place in the AIM 2020 Video Frame Interpolation Challenge, 2nd place in the NTIRE 2021 HDR Enhancement Challenge, 3rd place in the UDC 2020 Under-Display Camera Restoration Challenge. I serve as a reviewer for various top journals and conferences, including TPAMI, TIP, TCSVT, TMM, CVPR, ICCV, ECCV, NeurIPS, etc.

Current Research Focus

My current research focuses on pioneering a new generation of multimodal foundation models that integrate generation and understanding within a unified architecture. Specifically:

  • Unified Multimodal Architectures: Designing new-generation frameworks (e.g., discrete diffusion, autoregressive hybrids) that integrate text, image, video, and audio tasks, enabling coherent cross-modal representation, reasoning, and generation.
  • Knowledge-Driven and Causality-Aware Modeling: Embedding structured world knowledge, physical realism, and causal reasoning into multimodal models, moving beyond perceptual fidelity toward scientifically grounded and logically consistent outputs.
  • General Low-Level Vision Models: Consolidating diverse low-level vision tasks — restoration, enhancement, style transfer, and dense prediction — into a robust multimodal framework, advancing detail recovery, fidelity, and generalization for real-world applications.
  • Post-training and Reward Alignment: Developing multimodal alignment and reinforcement learning paradigms, incorporating human preference modeling and expert feedback, to ensure outputs that are not only high-quality and aesthetic but also reliable, interpretable, and scientifically valid.

I am open to collaboration and discussions. Feel free to reach out at liuyihao@pjlab.org.cn or liuyihao14@mails.ucas.ac.cn

news

Jul 10, 2026 One paper accepted by ACM MM’26. MIGM-Shortcut accelerates masked image generation by learning latent controlled dynamics: instead of running the heavy base model at every sampling step, it uses a lightweight shortcut model to predict feature evolution from previous features and newly sampled tokens. Applied to MaskGIT and Lumina-DiMOO, MIGM-Shortcut greatly improves the quality-speed trade-off and achieves over 4x acceleration on Lumina-DiMOO while maintaining generation quality. [Homepage] [Paper].
Jun 18, 2026 Three papers accepted by ECCV’26.
May 04, 2026 I’m glad to share our ICML 2026 work StableI2I, a fidelity-oriented evaluation framework for image-to-image generation. Rather than only asking whether an edited/restored image looks good or follows the instruction, StableI2I focuses on what has been unintentionally changed. It jointly considers the input image, output image, and I2I instruction to diagnose content drift across semantic consistency, structural fidelity, and low-level appearance, covering errors such as object addition/removal/replacement, repainting, misalignment, noise, blur, and color cast. We release StableI2I-Bench, together with StableI2I and StableI2I-PLUS models, to support fine-grained I2I fidelity diagnosis and scoring for more faithful and controllable image editing/restoration systems. [Homepage] [GitHub] [StableI2I-Bench] [StableI2I Model] [StableI2I-PLUS] [Paper].
May 01, 2026 Five papers accepted by ICML’26. UniPercept was selected as a Spotlight paper.
Feb 21, 2026 Four papers accepted by CVPR’26.
Jan 27, 2026 Four papers accepted by ICLR’26.
Dec 30, 2025 I’m happy to share our new work UniPercept, which tackles a key blind spot of today’s multimodal LLMs: perceptual-level image understanding — how images look and feel to humans — covering aesthetics, quality, structure, and texture. Our release includes UniPercept-Bench, a unified benchmark spanning IAA/IQA/ISTA and supporting both Visual Rating (VR) and Visual Question Answering (VQA) evaluations. We also introduce the UniPercept baseline model to generalize across VR and VQA settings. Beyond benchmarking, UniPercept can be used as a reward model for post-training text-to-image systems and as a perceptual diagnostic tool for analyzing model outputs and datasets. [Homepage] [GitHub] [ UniPercept-Bench] [ UniPercept Model] [Paper].
Oct 21, 2025 We present PICABench, a new benchmark and evaluation protocol for assessing physical realism in image editing — an often overlooked dimension in current generative models. PICABench systematically evaluates the physical consequences across eight sub-dimensions spanning optics, mechanics, and state transitions, with a reliable PICAEval protocol combining VLM-as-a-judge and region-level human annotations. We also build PICA-100K, a dataset for learning physics from videos. Evaluations show that physical realism remains a major challenge. PICABench aims to drive the next wave of physics-aware, causally consistent image editing. [Homepage] [GitHub] [ PICABench Dataset] [ PICA-100K Dataset] [Paper].

selected publications

2026

  1. ACM MM
    lcd.png
    Accelerating Masked Image Generation by Learning Latent Controlled Dynamics
    Kaiwen Zhu, Quansheng Zeng, Yuandong Pu, Shuo Cao, Xiaohui Li, Yi Xin, Qi Qin, Jiayang Li, Yu Qiao, Jinjin Gu, and Yihao Liu†
    arXiv preprint arXiv:2602.23996, 2026
    arXiv Code
  2. ICML
    stablei2i.png
    StableI2I: Spotting Unintended Changes in Image-to-Image Transition
    Jiayang Li, Shuo Cao, Xiaohui Li, Zhizhen Zhang, Kaiwen Zhu, Yule Duan, Yu Qiao, Jian Zhang, and Yihao Liu†
    arXiv preprint arXiv:2605.04453, 2026
    arXiv Code Website
  3. arXiv
    cfworld.png
    Are Text-to-Image Models Inductivist Turkeys? A Counterfactual Benchmark for Causal Reasoning
    Jiayi Lei, Yuandong Pu, Xingyu Han, Rongpeng Zhu, Jing Xu, Jinyao Wang, Zijian Zhou, Bin Fu, Yuewen Cao†, Yihao Liu†, and Hongsheng Li†
    arXiv preprint arXiv:2606.24548, 2026
    arXiv Code Website
  4. arXiv
    fepbench.png
    Faithful, Enriched, and Precise: Benchmarking Natural-Science Illustration Generation by T2I models
    Yifan Chang, Jiaxin Ai, Jianwen Sun, Yuandong Pu, Siqi Luo, Liangliang Zhao, Yuchen Ren, Minghao Liu, Yunfei Yu, Yu Qiao, Kaipeng Zhang†, and Yihao Liu†
    arXiv preprint arXiv:2606.05949, 2026
    arXiv
  5. arXiv
    sketch-then-paint.png
    Sketch Then Paint: Hierarchical Reinforcement Learning for Diffusion Multi-Modal Large Language Models
    Siqi Luo, Jianghan Shen, Yi Xin, Huayu Zheng, Haoxing Chen, Yan Tai, Yue Li, Junjun He, Yihao Liu, Guangtao Zhai, Yuewen Cao, and Xiaohong Liu
    arXiv preprint arXiv:2605.16842, 2026
    arXiv
  6. arXiv
    massive-blur-priors.png
    Toward Generalizable Deblurring: Leveraging Massive Blur Priors with Linear Attention for Real-World Scenarios
    Yuanting Gao, Shuo Cao, Xiaohui Li, Yuandong Pu, Yihao Liu†, and Kai Zhang†
    arXiv preprint arXiv:2601.06525, 2026
    arXiv Code Website

2025

  1. ECCV
    onevae.png
    OneVAE: Joint Discrete and Continuous Optimization Helps Discrete Video VAE Train Better
    Yupeng Zhou, Zhen Li, Ziheng Ouyang, Yuming Chen, Ruoyi Du, Daquan Zhou, Bin Fu, Yihao Liu, Peng Gao, Ming-Ming Cheng, and  others
    arXiv preprint arXiv:2508.09857, 2025
    arXiv Code
  2. ECCV
    omnilv.png
    Lumina-omnilv: A unified multimodal framework for general low-level vision
    Yuandong Pu, Le Zhuo, Kaiwen Zhu, Liangbin Xie, Wenlong Zhang, Xiangyu Chen, Peng Gao, Yu Qiao, Chao Dong, and Yihao Liu†
    arXiv preprint arXiv:2504.04903, 2025
    arXiv Code Website
  3. ICML
    unipercept.png
    UniPercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture
    Shuo Cao, Jiayang Li, Xiaohui Li, Yuandong Pu, Kaiwen Zhu, Yuanting Gao, Siqi Luo, Yi Xin, Qi Qin,  ..., and Yihao Liu†
    arXiv preprint arXiv:2512.21675, 2025
    arXiv
  4. CVPR
    dmlLm-tts.png
    dMLLM-TTS: Self-Verified and Efficient Test-Time Scaling for Diffusion Multi-Modal Large Language Models
    Yi Xin, Siqi Luo, Tianxiang Xu, Qi Qin, Haoxing Chen, Kaiwen Zhu, Zhiwei Zhang, Yangfan He, Rongchao Zhang, Jinbin Bai, Shuo Cao, Bin Fu, Junjun He, Yihao Liu, Yuewen Cao, and Xiaohong Liu
    arXiv preprint arXiv:2512.19433, 2025
    arXiv Code
  5. CVPR
    flashvsr.png
    FlashVSR: Towards Real-Time Diffusion-Based Streaming Video Super-Resolution
    Junhao Zhuang, Shi Guo, Xin Cai, Xiaohui Li, Yihao Liu, Chun Yuan, and Tianfan Xue
    arXiv preprint arXiv:2510.12747, 2025
    arXiv Code Website
  6. CVPR
    artimuse.png
    Artimuse: Fine-grained image aesthetics assessment with joint scoring and expert-level understanding
    Shuo Cao, Nan Ma, Jiayang Li, Xiaohui Li, Lihao Shao, Kaiwen Zhu, Yu Zhou, Yuandong Pu, Jiarui Wu, Jiaquan Wang, Bo Qu, Wenhai Wang, Yu Qiao, Dajuin Yao, and Yihao Liu†
    arXiv preprint arXiv:2507.14533, 2025
    arXiv Code Website
  7. ICLR
    omniweather.png
    Omni-Weather: Unified Multimodal Foundation Model for Weather Generation and Understanding
    Zhiwang Zhou, Yuandong Pu, Xuming He, Yidi Liu,  ..., Yihao Liu†, Wenlong Zhang†, and Lei Bai†
    arXiv preprint arXiv:2512.21643, 2025
    arXiv
  8. ICLR
    picabench.png
    PICABench: How Far Are We from Physically Realistic Image Editing?
    Yuandong Pu, Le Zhuo, Songhao Han, Jinbo Xing, Kaiwen Zhu, Shuo Cao, Bin Fu, Si Liu, Hongsheng Li, Yu Qiao, Wenlong Zhang, Xi Chen, and Yihao Liu†
    arXiv preprint arXiv:2510.17681, 2025
    arXiv Code Website
  9. ICLR
    linearsr.png
    LinearSR: Unlocking Linear Attention for Stable and Efficient Image Super-Resolution
    Xiaohui Li, Shaobin Zhuang, Shuo Cao, Yang Yang, Yuandong Pu, Qi Qin, Siqi Luo, Bin Fu, and Yihao Liu†
    arXiv preprint arXiv:2510.08771, 2025
    arXiv
  10. ICLR
    factuality-structured-visuals.png
    Factuality Matters: When Image Generation and Editing Meet Structured Visuals
    Le Zhuo, Songhao Han, Yuandong Pu, Boxiang Qiu, Sayak Paul, Yue Liao, Yihao Liu, Jie Shao, Xi Chen, Si Liu, and Hongsheng Li
    arXiv preprint arXiv:2510.05091, 2025
    arXiv Code Website
  11. AAAI
    synweather.png
    SynWeather: Weather Observation Data Synthesis across Multiple Regions and Variables via a General Diffusion Transformer
    Kaiyi Xu, Junchao Gong, Zhiwang Zhou, Zhangrui Li, Yuandong Pu, Yihao Liu, Ben Fei, Fenghua Ling, Wenlong Zhang, and Lei Bai
    arXiv preprint arXiv:2511.08291, 2025
    arXiv Code Website
  12. ICCV
    diffvsr.png
    DiffVSR: Revealing an Effective Recipe for Taming Robust Video Super-Resolution Against Complex Degradations
    Xiaohui Li*, Yihao Liu*†, Shuo Cao, Ziyan Chen, Shaobin Zhuang, Xiangyu Chen, Yinan He, Yi Wang, and Yu Qiao
    arXiv preprint arXiv:2501.10110, 2025
    arXiv Code Website
  13. TIP
    l2s.png
    Learning to see low-light images via feature domain adaptation
    Qirui Yang, Qihua Cheng, Huanjing Yue, Le Zhang, Yihao Liu, and Jingyu Yang
    IEEE Transactions on Image Processing, 2025
  14. arXiv
    dimoo.png
    Lumina-dimoo: An omni diffusion large language model for multi-modal generation and understanding
    Yi Xin, Qi Qin, Siqi Luo, Kaiwen Zhu, Juncheng Yan, Yan Tai, Jiayi Lei, Yuewen Cao, Keqi Wang, Yibin Wang,  ..., and Yihao Liu†
    arXiv preprint arXiv:2510.06308, 2025
    arXiv Code Website
  15. arXiv
    dualx-vsr.png
    DualX-VSR: Dual Axial Spatial\timesTemporal Transformer for Real-World Video Super-Resolution without Motion Compensation
    Shuo Cao, Yihao Liu, Xiaohui Li, Yuanting Gao, Zhou Yu, and Chao Dong
    arXiv preprint arXiv:2506.04830, 2025
    arXiv

2024

  1. ICLR
    weathergfm.png
    WeatherGFM: Learning A Weather Generalist Foundation Model via In-context Learning
    Xiangyu Zhao, Zhiwang Zhou, Wenlong Zhang, Yihao Liu, Xiangyu Chen, Junchao Gong, Hao Chen, Ben Fei, Shiqi Chen, Wanli Ouyang, Xiao-Ming Wu, and Lei Bai
    arXiv preprint arXiv:2411.05420, 2024
    arXiv
  2. ACM MM
    genlv.png
    Learning A Low-Level Vision Generalist via Visual Task Prompt
    Xiangyu Chen, Yihao Liu, Yuandong Pu, Wenlong Zhang, Jiantao Zhou, Yu Qiao, and Chao Dong
    In Proceedings of the 32nd ACM International Conference on Multimedia, 2024
    Code
  3. ECCV
    grids.png
    GRIDS: Grouped Multiple-Degradation Restoration with Image Degradation Similarity
    Shuo Cao*, Yihao Liu*, Wenlong Zhang, Yu Qiao, and Chao Dong
    In European Conference on Computer Vision, 2024
  4. ECCV
    x-restormer.png
    A Comparative Study of Image Restoration Networks for General Backbone Network Design
    Xiangyu Chen, Zheyuan Li, Yuandong Pu, Yihao Liu, Jiantao Zhou, Yu Qiao, and Chao Dong
    In European Conference on Computer Vision, 2024
    Code
  5. ICML
    promptgip.png
    Unifying Image Processing as Visual Prompting Question Answering
    Yihao Liu*, Xiangyu Chen*, Xianzheng Ma*, Xintao Wang, Jiantao Zhou, Yu Qiao, and Chao Dong
    In Proceedings of the 41st International Conference on Machine Learning (ICML), 2024
    Code
  6. CVMJ
    tcvc.gif
    Temporally Consistent Video Colorization with Deep Feature Propagation and Self-Regularization Learning
    Yihao Liu*, Hengyuan Zhao*, Kelvin CK Chan, Xintao Wang, Chen Change Loy, Yu Qiao, and Chao Dong
    Computational Visual Media, 2024
    Code
  7. arXiv
    dpr-tonemapping.png
    Learning Differential Pyramid Representation for Tone Mapping
    Qirui Yang, Yinbo Li, Yihao Liu, Peng-Tao Jiang, Fangpu Zhang, Qihua Cheng, Huanjing Yue, and Jingyu Yang
    arXiv preprint arXiv:2412.01463, 2024
    arXiv Website

2023

  1. TPAMI
    srga.png
    Evaluating the Generalization Ability of Super-Resolution Networks
    Yihao Liu, Hengyuan Zhao, Jinjin Gu, Yu Qiao, and Chao Dong
    IEEE Transactions on pattern analysis and machine intelligence, 2023
    Code
  2. TPAMI
    cp3.png
    CP3: Unifying Point Cloud Completion by Pretrain-Prompt-Predict Paradigm
    Mingye Xu, Yali Wang, Yihao Liu, Tong He, and Yu Qiao
    IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023
  3. CVPR
    degae.png
    DegAE: A New Pretraining Paradigm for Low-Level Vision
    Yihao Liu, Jingwen He, Jinjin Gu, Xiangtao Kong, Yu Qiao, and Chao Dong
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023
    Code
  4. CVPR
    maskeddenoising.png
    Masked Image Training for Generalizable Deep Image Denoising
    Haoyu Chen, Jinjin Gu, Yihao Liu, Salma Abdel Magid, Chao Dong, Qiong Wang, Hanspeter Pfister, and Lei Zhu
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023
    Code

2022

  1. TMM
    csrnet-l.png
    Very Lightweight Photo Retouching Network with Conditional Sequential Modulation
    Yihao Liu*, Jingwen He*, Xiangyu Chen, Zhengwen Zhang, Hengyuan Zhao, Chao Dong, and Yu Qiao
    IEEE Transactions on Multimedia, 2022
    Code
  2. TPAMI
    bsr_survey.png
    Blind Image Super-Resolution: A Survey and Beyond
    Anran Liu, Yihao Liu, Jinjin Gu, Yu Qiao, and Chao Dong
    IEEE transactions on pattern analysis and machine intelligence, 2022

2021

  1. TPAMI
    ranksrgan-j.png
    RankSRGAN: Super Resolution Generative Adversarial Networks with Learning to Rank
    Wenlong Zhang, Yihao Liu, Chao Dong, and Yu Qiao
    IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021
    Code
  2. TPAMI
    cresmd.png
    Interactive Multi-Dimension Modulation for Image Restoration
    Jingwen He, Chao Dong, Yihao Liu, and Yu Qiao
    IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021
    Code
  3. ICCV
    l2m.png
    Learn to Match: Automatic Matching Network Design for Visual Tracking
    Zhipeng Zhang, Yihao Liu, Xiao Wang, Bing Li, and Weiming Hu
    In International Conference on Computer Vision (ICCV), 2021
    Code
  4. arXiv
    ddr.png
    Discovering" Semantics" in Super-Resolution Networks
    Yihao Liu*, Anran Liu*, Jinjin Gu, Zhipeng Zhang, Wenhao Wu, Yu Qiao, and Chao Dong
    arXiv preprint arXiv:2108.00406, 2021
    arXiv

2020

  1. ECCV
    csrnet.png
    Conditional Sequential Modulation for Efficient Global Image Retouching
    Jingwen He*, Yihao Liu*, Yu Qiao, and Chao Dong
    In European Conference on Computer Vision (ECCV), 2020
  2. ECCVW
    eqvi.png
    Enhanced Quadratic Video Interpolation
    Yihao Liu*, Liangbin Xie*, Li Siyao, Wenxiu Sun, Yu Qiao, and Chao Dong
    In European Conference on Computer Vision (ECCV) Workshops, 2020
    Code
  3. AAAI
    fdgan.png
    FD-GAN: Generative Adversarial Networks with Fusion-Discriminator for Single Image Dehazing
    Yu Dong*, Yihao Liu*, He Zhang, Shifeng Chen, and Yu Qiao
    In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2020

2019

  1. ICCV
    RankSRGAN: Generative Adversarial Networks with Ranker for Image Super-Resolution
    Wenlong Zhang, Yihao Liu, Chao Dong, and Yu Qiao
    In International Conference on Computer Vision (ICCV), 2019
    Code

2018

  1. ECCVW
    ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks
    Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Yu Qiao, and Chen Change Loy
    In Proceedings of the European conference on computer vision (ECCV) workshops, 2018
    Code
Flag Counter